WO2023017901A1

WO2023017901A1 - Breast cancer risk assessment system and method

Info

Publication number: WO2023017901A1
Application number: PCT/KR2021/013806
Authority: WO
Inventors: 성주헌; 차유현; 안주영; 정여진; 이종원
Original assignee: 서울대학교산학협력단
Priority date: 2021-08-12
Filing date: 2021-10-07
Publication date: 2023-02-16
Also published as: US20240186015A1; KR20230025555A

Abstract

A breast cancer risk assessment method of a breast cancer risk assessment system using a mammography image, according to one embodiment of the present invention, comprises steps in which the system: generates assessment data including breast density information, which is generated by measuring, from a mammography image of a breast to be assessed, the density of the breast to be assessed, and the breast pattern information, which is generated by extracting, from the mammography image, a feature pattern of the breast to be assessed; and calculates the risk of breast cancer for the breast to be assessed by applying a preset weight to respective pieces of information included in the assessment data.

Description

Breast cancer risk assessment system and method

The present invention relates to a breast cancer risk assessment system and method, and more particularly, to a system and method for evaluating breast cancer risk by measuring breast density and extracting characteristic patterns of breast cancer incidence.

Breast cancer is known to be one of the most common cancers in women along with thyroid cancer in Korea. According to the data of the Central Cancer Registry published in 2017, breast cancer occurred in 19,219 cases in 2017 alone, accounting for 9.0% of all cancers in both men and women.

Breast density, which can be measured through mammograms for breast cancer detection, is known as an indicator that can predict the occurrence of breast cancer. According to the results of a study conducted on Korean women, it has been reported that women in the upper quartile of breast density have a two to four times higher risk of breast cancer compared to women in the lower quartile. Therefore, if a screening method for breast cancer high-risk group considering breast density can be implemented in the breast cancer screening process, a customized prevention strategy for breast cancer high-risk group can be established to contribute to breast cancer prevention through early detection of breast cancer and voluntary environmental improvement.

Currently, a conventional method known to be able to most accurately evaluate breast density is a semi-automated measurement method by an expert assisted by a computer program. In the conventional case, although it is possible to accurately evaluate breast density, there is a disadvantage in that it is labor-intensive because an expert visually reads the breast density. Therefore, there is a need for a method for automatically evaluating breast density for cost-effective breast density measurement. In addition, in the prior art, since a breast density measurement value varies depending on technical conditions such as a radiation dose and an imaging device manufacturer, there is a problem in which reliability is low.

SUMMARY OF THE INVENTION The present invention is intended to solve the above problems, and one technical problem is to provide a system and method for automating breast cancer risk assessment through breast density measurement.

In addition, another technical task of the present invention is to perform multi-level breast density evaluation and pattern analysis using machine-learning based artificial intelligence technology, and to provide an image-based breast cancer risk evaluation system based thereon. .

In addition, another technical task of the present invention is to provide a breast cancer risk assessment system that integrates image-based breast cancer risk assessment results and clinical information-based risk assessment results.

The technical problems to be achieved by the present invention are not limited to the above technical problems, and other technical problems of the present invention can be derived from the following description.

As a technical means for solving the above-described technical problem, a breast cancer risk evaluation method of a breast cancer risk evaluation system using a mammogram image according to a first aspect of the present invention includes: generating evaluation data including breast density information generated by measuring the density of the breast to be evaluated and breast pattern information generated by extracting a characteristic pattern of the breast to be evaluated from the mammography image; and calculating a risk of breast cancer occurrence for the breast to be evaluated by applying a predetermined weight to each piece of information included in the evaluation data.

In addition, the breast cancer risk evaluation system using mammography images according to the second aspect of the present invention includes a communication module for receiving the mammography images, a memory for storing a breast cancer risk evaluation program, and a breast cancer risk evaluation program stored in the memory. includes a processor that executes The processor executes the breast cancer risk evaluation program to extract breast density information generated by measuring the density of the breast to be evaluated from a mammogram image of the breast to be evaluated, and a characteristic pattern of the breast to be evaluated from the mammogram image. generating evaluation data including breast pattern information generated by doing so, and calculating a risk of breast cancer for the breast to be evaluated by applying a predetermined weight to each piece of information included in the evaluation data. .

According to the present invention, it is possible to automate breast density evaluation by constructing a highly reliable database based on mammography image data and image analysis information (tagged image database) and applying a machine-learning method to the databased data. It is possible to automatically predict breast density by learning from vast amount of data, and extract a pattern for predicting the occurrence of breast cancer from mammography image data using patient-normal decision result information.

In addition, according to the present invention, multi-level breast density evaluation and pattern analysis can be performed using machine-learning based artificial intelligence technology, and risk evaluation based on image-based breast cancer risk evaluation results and clinical information. Results can be combined to assess breast cancer risk.

In addition, according to the present invention, it is possible to extract features related to breast cancer risk through weight learning by applying a machine-learning method based on a tagged image database, and thus, a pattern directly related to breast cancer occurrence can be extracted. Therefore, it is possible to more accurately assess the risk of breast cancer.

In addition, according to the present invention, it is possible to automate the breast density evaluation, which was visually evaluated by trained experts, and to calculate more indicators (multilevel breast density, pattern analysis, etc.) from breast images compared to the prior art. A breast cancer risk assessment method having high efficiency at a relatively low cost compared to the prior art can be implemented.

In addition, according to the present invention, breast cancer risk information including mammography image information can be provided to a examinee in real time during a breast cancer screening process, and the high-risk group and low-risk group can be classified accordingly to contribute to the efficiency of screening resources and the prevention of breast cancer. .

The effects of the present invention are not limited to the effects described above, and include all effects understood from the following description.

1 is a block diagram showing the configuration of a breast cancer risk assessment system according to an embodiment of the present invention.

2 to 6 are diagrams for explaining a method of analyzing a mammography image according to an embodiment of the present invention.

7 is a flowchart illustrating steps of a breast cancer risk assessment method according to an embodiment of the present invention.

8 and 9 are flowcharts illustrating detailed procedures for some steps of the breast cancer risk assessment method shown in FIG. 7 .

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. However, the present invention may be implemented in many different forms, and is not limited to the embodiments described herein. In addition, the accompanying drawings are only for easy understanding of the embodiments disclosed in this specification, and the technical ideas disclosed in this specification are not limited by the accompanying drawings. In order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and the size, shape, and shape of each component shown in the drawings may be variously modified. Same/similar reference numerals are assigned to the same/similar parts throughout the specification.

The suffixes "module" and "unit" for components used in the following description are given or used interchangeably in consideration of ease of writing the specification, and do not have meanings or roles that are distinct from each other by themselves. In addition, in describing the embodiments disclosed in this specification, if it is determined that a detailed description of related known technologies may obscure the gist of the embodiments disclosed in this specification, the detailed description is omitted.

Throughout the specification, when a part is said to be “connected (connected, contacted, or combined)” with another part, this is not only the case where it is “directly connected (connected, contacted, or coupled)”, but also has other members in the middle. It also includes the case of being "indirectly connected (connected, contacted, or coupled)" between them. In addition, when a part "includes (provides or provides)" a certain component, it does not exclude other components, but "includes (provides or provides)" other components unless otherwise specified. means you can

Terms indicating ordinal numbers such as first and second used in this specification are used only for the purpose of distinguishing one element from another, and do not limit the order or relationship of elements. For example, a first element of the present invention may be termed a second element, and similarly, the second element may also be termed a first element.

1 is a block diagram showing the configuration of a breast cancer risk assessment system 100 according to an embodiment of the present invention. The breast cancer risk assessment system 100 may be implemented as a server or a computing device, and may operate in a cloud computing service model such as Software as a Service (SaaS), Platform as a Service (PaaS) or Infrastructure as a Service (IaaS). can In addition, the breast cancer risk assessment system 100 may be constructed in the form of a private cloud, public cloud, or hybrid cloud system, but the scope of the present invention is not limited thereto. Although not shown in the drawing, the breast cancer risk assessment system 100 may transmit/receive information with a medical imaging device and a database in which mammography images are stored. The breast cancer risk evaluation system 100 may calculate the risk of breast cancer using a mammogram image captured by a medical imaging device and a database in which mammogram images are converted into big data and stored.

Referring to FIG. 1 , a breast cancer risk assessment system 100 includes a communication module 110 that transmits and receives information with an external device and receives a mammography image, a memory 120 storing a breast cancer risk assessment program, and a memory 120. ) and a processor 130 that executes a breast cancer risk assessment program stored in. The name of the breast cancer risk assessment program is set for convenience of explanation, and the name itself does not limit the function of the program, and may be set to various program names. The processor 130 executes a program to receive mammography images, evaluate and calculate breast density from the images, analyze breast patterns, set density areas based on image pixel brightness values, and based on personal information of a person with breasts to be evaluated. Based on clinical information settings, breast density information, breast pattern information, and clinical information, the risk of the breast to be evaluated can be calculated.

The above-described communication module 110 may include a device including hardware and software necessary for transmitting and receiving signals such as control signals or data signals with other network devices through wired or wireless connections. The memory 120 should be interpreted as collectively referring to a non-volatile storage device that continuously maintains stored information even when power is not supplied and a volatile storage device that requires power to maintain stored information. In addition, the memory 120 may perform a function of temporarily or permanently storing data, and may be a magnetic storage medium or flash storage medium in addition to a volatile storage device that requires power to maintain stored information. media), but the scope of the present invention is not limited thereto. The processor 130 may include various types of devices that control and process data. Also, the processor 130 may refer to a data processing device embedded in hardware having a physically structured circuit to perform functions expressed by codes or instructions included in a program. In one example, the processor 130 may include a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated circuit (ASIC), an FPGA ( field programmable gate array), etc., but the scope of the present invention is not limited thereto. The terminal is, for example, a laptop, desktop, and laptop equipped with a web browser, and all types of handhelds such as smartphones and tablet PCs as wireless communication devices that ensure portability and mobility. handheld)-based wireless communication device. In addition, the network may include a wired network such as a Local Area Network (LAN), a Wide Area Network (WAN) or a Value Added Network (VAN), a mobile radio communication network, a satellite communication network, and the like. It can be implemented in all kinds of wireless networks such as

In more detail, the processor 130 may execute the breast cancer risk assessment program stored in the memory 120 to perform the following functions and procedures. The processor 130 measures the density of the breast to be evaluated from a mammography image of the breast to be evaluated and generates breast density information. The processor 130 generates evaluation data including breast pattern information generated by extracting a characteristic pattern of a breast to be evaluated from a breast imaging image. The processor 130 calculates the risk of breast cancer occurrence for the breast to be evaluated by applying a predetermined weight to each piece of information included in the evaluation data. Here, the mammography image may be a DICOM (Digital Imaging and Communications in Medicine) type image. The feature pattern may include an abnormal breast feature pattern of the breast to be evaluated.

In addition, the processor 130 may execute a breast cancer risk assessment program to generate clinical information based on at least one or more of age information, genome information, and breast cancer family history information of a specific person corresponding to the breast to be evaluated. Clinical information may include physical information, female history, lifestyle, family history, and the like of a person having a breast to be evaluated. The processor 130 may perform preprocessing to adjust the mammography image to have a preset contrast and size structure. The processor 130 may calculate the size of the breast to be evaluated by segmenting the breast from the preprocessed mammography image. The processor 130 may calculate the breast density using the pixel density of the breast area.

Furthermore, the processor 130 executes a breast cancer risk assessment program to first density, second density, and third density in descending order of brightness values according to a predetermined criterion of brightness values for each pixel in the mammography image. can be divided into areas. At this time, the processor 130 may calculate the sizes of the first to third density regions of the breast. The size of the breast may be an area occupied by the breast in the image or an area of the breast. Also, the processor 130 may calculate the breast density for each region using pixel densities in the first to third density regions.

In one example, the size of the breast region for each of the first to third density regions is determined by an absolute breast size value corresponding to an area value of the breast region for each region of the first to third density regions and the first to third densities of the breast region. A breast size relative value corresponding to a value obtained by dividing the area value of a region by the total area value of the breast may be included.

The processor 130 may perform breast cancer risk assessment using an artificial intelligence model using deep learning technology. In one example, the processor 130 executes a breast cancer risk evaluation program, performs vector transformation to which a weight is applied based on a plurality of mammography images to be learned, calculates a breast density of an image breast region, and calculates a breast density and A trained artificial intelligence model may be created to minimize a difference in breast density derived by an expert targeting mammography images to be learned. The processor 130 may measure the density of the breast to be evaluated by receiving a medical imaging image of the breast to be evaluated using the artificial intelligence model. In another example, the processor 130 executes a breast cancer risk assessment program, performs machine learning based on mammogram images of a normal person and a breast cancer patient, and extracts a feature pattern of an abnormal breast from an input image. Intelligence models can be created. In this case, the processor 130 may extract an abnormal breast feature pattern of the breast to be evaluated from the mammography image using the artificial intelligence model.

Referring to FIGS. 2 and 3 together with FIG. 1 , the criterion, basis, and setting method for dividing the first to third density regions in the captured image of the breast to be evaluated are as follows. The processor 130 may evaluate the subject's breast density in pixel units through three stages. This method can be referred to as a multi-level breast density evaluation method.

For example, breast density may be classified into three levels of normal density, high density, and ultra-high density by the processor 130 . The processor 130 may express the multilevel breast density as an absolute amount (cm^2) and a relative amount (%). The absolute amount is a value obtained by converting the absolute amount of breast parenchymal tissue into cm^2, and the relative amount is a value expressed as a percentage (%) by dividing the absolute amount by the total breast area.

As shown in FIG. 2 , in the case of a normal density, the processor 130 may set a portion 21 that is regarded as parenchymal tissue on a mammogram image, and in the case of a high density, a gray portion included in the existing normal density may be excluded. It can be set to the bright area 22, and in the case of ultra-high density, it can be set to the brightest area 23 on the image. A standard of brightness may be preset. In particular, the higher the density, the higher the predictive power of breast cancer. Therefore, although the overall ratio is low, ultra-high density best reflects the risk of breast cancer, and it is not related to general density, so it can be used as an independent indicator. Accordingly, the processor 130 may highly evaluate the risk of the breast to be evaluated corresponding to a mammography image having many ultra-high density regions.

In the tagged image database 31 storing mammography images in FIG. 3 , part 21 may be set to the first density, part 22 may be set to the second density, and part 23 may be set to the third density. can Values of the first to third densities may be values derived by measuring density values for each density area. In the tagged image database 31, the preprocessing image database 32 and the segmented image database 33 may be obtained through the preprocessing process and the image segmentation process, respectively.

Examples of main procedures performed by the processor 130 according to the above-described embodiment of the present invention are as follows.

The processor 130 may perform image-based risk assessment that evaluates risk from mammography images based on a machine-learning method and clinical information-based risk assessment that evaluates risk by applying a statistical method to clinical and genomic information. . Image-based risk assessment consists of a breast density evaluation process that automatically evaluates the breast density of an input image through a weighted variable, and a pattern analysis process that extracts patterns related to breast cancer risk from the input image through a weighted variable and evaluates the risk through this process. can include

The processor 130 may calculate an integrated breast cancer risk by integrating the risk evaluation results of the image-based risk evaluation and the clinical information-based risk evaluation. The breast density evaluation process is a process of automatically evaluating breast density from an input image, and includes a breast segmentation process of segmenting a breast region, a preprocessing process of resizing an input image and enhancing local contrast, and and a dense region prediction process of predicting a dense region by performing vector transformation on the breast region, and a breast density prediction process of calculating a breast density by synthesizing the results of the breast segmentation process and the dense region prediction process.

A weight variable in the breast density evaluation process may be learned through a tagged image database representing images and image analysis information. In this learning process, an image preprocessed by a preprocessing process is used as an input image. The processor 130 divides the input image into training data and verification data, and repeats the process of adjusting the transform weight to minimize the cost function for the training data until the cost function for the verification data is minimized. Transformation weights can be adjusted in the region prediction process.

The processor 130 may output a breast density measurement value based on output results of the breast segmentation process and the dense region prediction process in the breast density prediction process. Breast density can be calculated for each density category by breast absolute density (dense area, DA) and density percentage (percent density, PD).

The pattern analysis process performed by the processor 130 may be learned through a tagged image database representing images and normal/patient determination information. In the case of the pattern analysis process, similar to the breast density evaluation process, the process of adjusting the transform weight to minimize the cost function by the learning process may include a process of repeating the process until the point at which the cost function for the verification data is minimized.

The image-based risk evaluation process performed by the processor 130 is a process of calculating the risk level by applying a conversion function that converts outputs of the breast density evaluation process and the pattern analysis process into a risk level.

The risk evaluation process based on clinical information performed by the processor 130 is a process of calculating a risk level according to clinical information and genetic information of a person having breasts subject to risk evaluation. This process involves dividing the subject into several phenotypes through genetic/family history information, and then squaring the relative risk estimated through the risk of breast cancer according to the age of each phenotype and clinical information. .

The integrated breast cancer risk assessment process performed by the processor 130 is a process of integrating the results of the image-based risk assessment process and the clinical information-based risk assessment process to calculate the final breast cancer risk. This process includes a process of calculating the risk level by applying a weighted average based on the normal/patient discrimination performance of each risk evaluation process for the tagged image database.

The breast density evaluation process performed by the processor 130 is a method of performing more accurate prediction by directly learning expert image analysis information. In the case of the prior art, if an expert performed an analysis of predicting breast density obtained by analyzing an image using a statistical method, in the present invention, a segmented image database can be built and learned from the expert's breast density measurement. . Therefore, since density evaluation for each pixel of an image is possible, accurate and multi-level evaluation of breast density is possible.

In the prior art, a method of calculating a breast cancer risk by calculating a predefined feature value from an image has been the main method. On the other hand, in the pattern analysis process performed by the processor 130, in this study, features related to breast cancer risk can be extracted through weight learning by applying a machine-learning method based on a tagged image database. Through this method, it is possible to more accurately evaluate the risk of breast cancer because it is possible to extract patterns directly related to the occurrence of breast cancer. In addition, according to an embodiment of the present invention, breast density evaluation, which has been visually evaluated by a trained expert, can be automated.

Hereinafter, a specific embodiment of the mammography risk assessment system 100 will be described with reference to FIGS. 4 and 5 together with FIG. 1 .

The breast map risk evaluation system 100 performs breast cancer risk evaluation through a breast density evaluation process, a pattern analysis process, a standard density conversion process, an image-based risk evaluation process, a clinical information-based risk evaluation process, and an integrated breast cancer risk evaluation process. can

The breast density evaluation process is a process of converting an input image into an output vector through a conversion vector, and then calculating and outputting the breast density (multilevel breast density) for each density category. The standard density conversion process is a process of receiving multi-level breast density, which is an output of the breast density evaluation process, as an input, converting the breast density for each category into a standard density standardized by age and body mass index, and then outputting the result. The pattern analysis process is a process of extracting a feature pattern related to breast cancer risk from an input image, performing risk analysis, and finally outputting the pattern risk.

The image-based risk assessment process is the process of applying a specific coefficient to the standard density, which is the output of the standard density conversion process, and the pattern risk, which is the output of the pattern analysis process, converting it into an image-based risk and then outputting it.

The breast density evaluation process includes a breast segmentation process, a preprocessing process, a dense region prediction process, and a breast density prediction process.

In the preprocessing process, an input image may be converted into a standard image size using bilinear interpolation, and then image normalization may be performed by applying contrast limited adaptive histogram equalization (CLAHE).

At this time, the image output by the preprocessing process is denoted as I(x,y). I(x,y) represents the pixel intensity of the image at the x, y coordinates of the image, and when the size of the standard image is w-by-h, it can be expressed as in Equation 1 below .

-Equation (1)-

Referring to FIG. 4 , in the dense region prediction process, a prediction segmented image 43 is output by applying a conversion function 42 to an input image 41 . An example of the transform function 42 in the dense region prediction process is a linear combination of an input image and a transform weight.

When the input image I(x,y) is composed of w-by-h pixels, the input image is expressed as a 2-dimensional vector (ranks of each dimension are w and h) as shown in Equation (2) below. can In Equation (2), xij means the (i,j)th pixel value.

-Equation (2)-

conversion function

As a specific example of , a conversion function F(x,y) as shown in Equation (3) below may be cited.

-Formula (3)-

The output vector T(X) obtained from the input image by the conversion function F(x, y) is a w-by-h vector and has the result of element-by-element multiplication of I(x,y) and F(x,y) as elements. . This can be expressed as Equation (4) below.

-Formula (4)-

The final output vector O(X) is obtained by creating a total of k T(x, y) through a total of k F(x, y), concatenating them, and applying the softmax function. can In this case, k is defined as the number of density categories, and the output vector O_c of category c, which is one of the k density categories, can be expressed as Equation (5) below.

-Equation (5)-

is the final output vector from the input image through the above conversion process

can be said to be a function that obtains conversion function

can use convolutional operation, which is an operation used in deep-learning implementation, considering complexity, and can use a combination of pooling operation, deconvolution, and nonlinear function. You will be able to.

The transform weight F used in the process of predicting the dense region performed by the processor 130 may be derived through a learning process as described above with reference to FIGS. 2 and 3, and this learning process is divided into the preprocessed image database 32. A process for learning a breast density evaluation process based on the image database 33 may be included.

The learning process may be performed by receiving batch data generated from the batch generator. The batch generator generates batch data from the pre-processing image database and the segmented image database, and transmits the batch data to the learning process so that the breast density evaluation process can be learned.

This learning process is a process for finding a transform weight set F having the smallest loss function. This process is performed by obtaining F that minimizes the cost function for the validation set by monitoring in the process of updating W using the change in weight (W) that minimizes the cost (loss) for the training set.

The learning process includes a first process of predicting the density of each pixel by applying the current transform weight to the input image, a second process of calculating a cost meaning a difference from the expert measurement value using a cost function, and a process of minimizing the cost. The transform weight is adjusted by repeating the third process of updating the transform weight by calculating the amount of change in the transform weight.

The learning process has a cost function, which can be expressed as cost(true_label, predicted label). A cross entropy function may be used as the cost function of the learning process, but is not limited thereto, and other cost functions may be set in various ways. In order to minimize the cost calculated by the cost function of the learning process, f_delta, the amount of change in the current transform weight, can be calculated and updated to f_new = f_old + f_delta.

In the process of minimizing the cost function for the training data by updating the transform weight by the learning process, the cost function for the verification data is monitored by the monitoring process. The update process is repeatedly performed until the point where the cost function for the verification data is minimized to determine the final weight.

Through the above-described learning process and monitoring process, a conversion weight that can best perform breast density evaluation can be learned. Finally, the transform function learned by the learning process

image

Apply the calculated vector to the output vector

, and the output vector can have a value between 0 and 1.

is the input image

It is a vector having a probability that a pixel of x, y coordinates belongs to each density category c as an element, and has the range of the following equation (6).

-Formula (6)-

output vector

In , the value that c, which means the density category, can have is an integer between 0 and k. This means that it is a detailed part. Prediction segmentation image, which is the final output result of the dense region prediction process

The expression to obtain is the following equation (7).

-Formula (7)-

In the breast image segmentation process performed by the processor 130, after going through the first process of converting the input image into a one-dimensional vector, a gaussian mixture model is applied to estimate the parameter, the value obtained by averaging (

) as a threshold, and is divided into a second process of segmenting a breast part from an input image.

When a threshold is applied to the input image in the second step of the breast segmentation process, the input image is converted into a binary breast segmentation image output S_breast, and the breast area (BA), which is the final output of the breast segmentation process, is converted to a breast segmentation image output S_breast. It is a value obtained by multiplying the total number of pixels divided into regions by the conversion factor K for conversion to cm^2.

The breast area, which is the output of the breast segmentation unit for the input image X, is calculated as shown in Equation (8) below.

-Formula (8)-

In the breast density estimation process of the breast density evaluation process performed by the processor 130, a final breast density prediction result is output by synthesizing results of the breast region and the dense region predicted from the breast segmentation process and the dense region prediction process.

The breast density output value according to the breast density prediction process is divided into absolute breast density (dense area, DA) and relative breast density (percent density, PD) for the density category.

The absolute breast density output from the input image by the breast density prediction process is the value obtained by multiplying the total number of pixels classified as dense regions by the conversion factor K for converting the unit to cm^2, and is the jth density for the input image X. degree of absolute breast density

can be obtained through Equation (9) below.

Ic(x)={1, x>=c

0, x<c}

-Formula (9)-

The relative breast density (PD) output from the input image by the breast density prediction process is a value obtained by converting the ratio of the absolute breast density area (DA) to the breast area (BA) output by the breast segmentation process into a percentage (%) am. Relative breast density of density category c for image X

is calculated through Equation (10) below.

-Formula (10)-

Standard density conversion process performed by the processor 130 This is a process of converting an output according to a breast density evaluation process into a standardized residual density. The implementation of the standard density transformation process is the first process of performing Box-cox transformation on the output (absolute breast density and relative breast density by density category) according to the breast density evaluation process, and the residual The second process of estimating the density, the third process of calculating the standardized residual density using μ and s, which mean the sample mean and standard deviation of the residual density obtained from expert measurements of the control data of the tagged image database in the results of the previous step include the process

When one of the outputs of the breast density evaluation process is x, the first to third processes of the standard density conversion process can be expressed as Equation (11) below.

- Formula (11) -

is the result of the box-cox transformation, and the transformation constant

is estimated through the maximum likelihood estimation method for the normal distribution with respect to the value predicted by the calibration variable while repeatedly performing conversion on the expert measurement values for the image of the control group in the tagged image database. .

In the above equation (11)

uses the coefficients estimated after fitting a linear regression model using a correction variable for the transformed β for the expert measurements of the control image in the tagged image database.

for the control image

After calculating , it is a value calculated as the mean and standard deviation.

The pattern analysis process performed by the processor 130 is a process of extracting a feature pattern related to breast cancer risk from an input image, performing a risk analysis, and finally outputting a pattern risk. Preprocessing, pattern extraction, and pattern risk prediction process includes

As described above, as an example of the preprocessing process, after converting an input image to a standard image size having a size of horizontal w and vertical h using linear interpolation, CLAHE (contrast limited adaptive histogram equalization) is applied Local contrast enhancement and normalization can be performed on images.

An example of implementation of the pattern extraction process performed by the processor 130 may include a convolutional neural network. Since the implementation of the convolutional neural network is a technique widely used in image processing techniques, a detailed description thereof will be omitted. It can be represented as a one-dimensional vector.

The pattern risk prediction process performed by the processor 130 is a process of receiving an output according to the pattern extraction process as an input and calculating a breast cancer risk using a conversion vector.

As an example of implementing the pattern risk prediction process, a neural network model (multi-layer perceptron) can be applied. As an example, a neural network model having one hidden layer will be described as an implementation example.

Let the output result from the pattern extraction process be a vector X with n elements, and when there is an n-by-m weight vector w(1) and a vector b(1) with m elements, the output vector h of the hidden layer is represented by the following equation (12).

-Equation (12)-

The pattern risk, which is an output result from the pattern extraction process, can be calculated by performing sigmoid transformation on the values obtained by applying the m-by-1 weight vectors w(2) and b(2) to the hidden layer output h. The formula for calculating the final pattern risk is Equation (13) below.

-Equation (13)-

Weights used in the pattern extraction process of the pattern analysis process performed by the processor 130 and the pattern risk prediction process may be learned through a learning process using a tagged image database. As described above, the learning process is performed by obtaining F that minimizes the cost function for the validation set in the process of finding the transform weight set F having the smallest value of the loss function.

The image-based risk assessment process performed by the processor 130 is a process of outputting an image-based risk rating by receiving outputs according to a breast density prediction process and a pattern analysis process as inputs. The output of the breast density prediction process for a total of k density categories

When the output of the pattern analysis process is P, the following equation (14) can be exemplified as an implementation of the image-based risk assessment process.

-Eq. (14)

in the above formula

Can be estimated from the above-described tagged image database.

The clinical information-based risk assessment process performed by the processor 130 is a process that is separate from the process applied to mammography images, and is a process of outputting a risk level considering clinical information and genome information of a subject.

There are various implementation examples of the clinical information-based risk assessment process, and in this embodiment, the Tyrer-Cuzick model, which is one widely used clinical model, will be described as an example. The risk calculation formula of the Tyrer-Cuzick model is shown in Equation (15) below.

-Equation (15)-

in the above formula

(BRCA gene absent/low penetrance gene absent), (BRCA gene absent/low penetrance gene present), (BRCA1 gene present/low penetrance gene absent), (BRCA1 gene present/low penetrance gene present), (BRCA2 gene present/ It is the probability of having the ith phenotype among 6 phenotypes.

in the above formula

is estimated through the patient's family history of breast/ovarian cancer.

is a predefined value as the probability that the ith phenotype has breast cancer between ages t1 and t2. in the above formula

is defined as the patient's relative risk of breast cancer after combining clinical information. The coefficients used in the above formula and the probability of breast cancer occurrence at a specific age are values calculated in advance based on the results of previous studies. In the case of the Tyrer-Cuzick model used in the clinical information-based risk assessment process, it is widely used to calculate the risk of breast cancer, so a detailed description is omitted.

The integrated breast cancer risk assessment process performed by the processor 130 is a process of calculating a breast cancer risk by combining an output according to an image-based risk assessment process and an output according to a clinical information-based risk assessment process. An example of implementing an integrated breast cancer risk assessment process is a weighted average using the performance of two predictive models. The weighted average can be expressed as Equation (16) below. The weights w_1 and w_2 of the integrated breast cancer evaluation process can be estimated from the above-described tagged image database.

-Equation (16)-

Referring to FIG. 6 , as an experimental example of the present invention based on the above descriptions, the processor 130 may perform breast cancer risk calculation based on a mammogram-based risk calculation automation algorithm. The corresponding algorithm includes breast density evaluation and pattern extraction as two major categories, and based on this, mammogram-based breast cancer risk can be output. In the experimental example, the processor 130 first performs preprocessing on a DICOM image as an input image. Next, the processor 130 divides the breast area to calculate the total area of the subject's breast. Next, the processor 130 calculates an absolute value (cm^2) for each multilevel breast density area and a relative value (%) for the total area. Next, the processor 130 may present evaluation information 61 including the final breast cancer risk by integrating the risk and clinical information calculated through the pattern analysis process in addition to the breast density.

7 is a flow chart showing steps of a breast cancer risk assessment method according to an embodiment of the present invention, and FIGS. 8 and 9 are flowcharts showing detailed procedures for some steps of the breast cancer risk assessment method shown in FIG. 7 . The breast cancer risk evaluation method according to this embodiment is a method using the breast cancer risk evaluation system 100 described above with reference to FIGS. 1 to 6 . Each step and detailed process of the breast cancer risk assessment method described below may be implemented through the above-described breast cancer risk assessment system 100 and may be performed by the processor 130 executing a program. Therefore, the contents of the embodiments described above with reference to FIGS. 1 to 6 may be equally applied to the following embodiments, and the contents overlapping with the above description will be omitted below.

Referring to FIG. 7 together with FIG. 1 , the breast cancer risk evaluation method using mammography images according to the present embodiment includes a step of generating evaluation data including breast density information and breast pattern information (S110) and based on the evaluation data and calculating the risk of breast cancer (S120).

In the step of generating evaluation data including breast density information and breast pattern information (S110), the breast cancer risk assessment system 100 measures the density of the breast to be evaluated from the mammography image of the breast to be evaluated, and generates breast density information and A step of generating evaluation data including breast pattern information generated by extracting a characteristic pattern of a breast to be evaluated from a mammography image. In the step of calculating the risk of breast cancer occurrence based on the evaluation data (S120), the breast cancer risk evaluation system 100 calculates the risk of breast cancer occurrence for the breast to be evaluated by applying a predetermined weight to each piece of information included in the evaluation data. It is a step to Here, the mammography image may be a DICOM (Digital Imaging and Communications in Medicine) type image.

In one example, generating evaluation data including breast density information and breast pattern information ( S110 ) may include a density measurement step using artificial intelligence technology. At this time, in the density measuring step, a vector transformation to which a weight is applied is performed based on a plurality of mammography images to be learned, to calculate the breast density of the breast part in the image, and the calculated breast density and the plurality of mammography images to be learned are targeted. It may be a step of measuring the density of the breast to be evaluated using an artificial intelligence model trained to minimize the difference in breast density of the breast region in the image drawn by the expert.

In one example, the step of generating evaluation data including breast density information and breast pattern information (S110) includes a pattern extraction step using artificial intelligence technology, and the feature pattern may include an abnormal breast feature pattern of the breast to be evaluated. . At this time, in the pattern extraction step, an abnormal breast pattern extraction artificial intelligence model configured to perform machine learning based on mammogram images of normal persons and breast cancer patients to extract feature patterns of abnormal breasts from the input images is used, and the mammogram images It may be a step of extracting an abnormal breast feature pattern of the breast to be evaluated from the breast.

In the step of generating evaluation data including breast density information and breast pattern information (S110), the breast cancer risk assessment system 100 provides information on at least one of age information, genetic information, and breast cancer family history information of a specific person corresponding to the breast to be evaluated. A step of generating clinical information generated based on may be further included. In this case, the evaluation data may include clinical information.

Referring to FIG. 8 , generating evaluation data including breast density information and breast pattern information (S110) includes pre-processing a mammogram image according to preset contrast and size standards (S111), and preprocessing the mammogram image. It may include calculating the size of the breast to be evaluated by dividing the breast region from (S112) and calculating the breast density using the pixel density of the breast region (S113). These steps may be density measurement steps. Calculating the breast density using the pixel density of the breast region ( S113 ) may include calculating the breast density for each region using pixel densities in the first to third density regions.

Referring to FIG. 9 , in step S112 of calculating the size of the breast to be evaluated by segmenting a breast region from a preprocessed mammography image (S112), according to a predetermined criterion of brightness values for each pixel in the mammography image, the brightness value is low. Sequentially classifying the first density region, the second density region, and the third density region (S112-1), and calculating the size of the first to third density regions of the breast (S112-2). can Here, the size of the breast region for each of the first to third density regions is the absolute value of the breast size corresponding to the area value of the first to third density regions of the breast region and the area value of the first to third density regions of the breast region. may include a breast size relative value corresponding to a value obtained by dividing ? by the total area value of the breast.

The breast cancer risk assessment method described above may be implemented in the form of a recording medium containing instructions executable by a computer, such as a program executed by a computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. Also, computer readable media may include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, programs or other data.

Those skilled in the art to which the present invention pertains will be able to understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention based on the above description. Therefore, the embodiments described above should be understood as illustrative in all respects and not limiting. The scope of the present invention is indicated by the following claims, and all changes or modifications derived from the meaning and scope of the claims and equivalent concepts should be interpreted as being included in the scope of the present invention.

The best mode for carrying out the above-described invention is as follows.

Since the present invention can be used in the medical industry related to disease diagnosis as a breast cancer diagnosis and evaluation technology, it has industrial applicability.

Claims

In the breast cancer risk assessment method of the breast cancer risk assessment system using mammography images,

a) Breast density information generated by the system by measuring the density of the breast to be evaluated from a mammogram image of the breast to be evaluated, and breast pattern information generated by extracting a characteristic pattern of the breast to be evaluated from the mammogram image Generating evaluation data comprising a; and

b) calculating, by the system, a risk of breast cancer occurrence for the breast to be evaluated by applying a predetermined weight to each piece of information included in the evaluation data.
According to claim 1,

In step a),

The system further comprises generating clinical information generated based on at least one or more of age information, genome information, and breast cancer family history information of a specific person corresponding to the breast to be evaluated,

The evaluation data includes the clinical information, breast cancer risk assessment method.
According to claim 1,

The mammography image is an image in the form of Digital Imaging and Communications in Medicine (DICOM), breast cancer risk assessment method.
According to claim 1,

Step a) includes a density measurement step,

The density measurement step,

a-1) pre-processing the mammography image according to preset contrast and size standards;

a-2) calculating the size of the breast to be evaluated by segmenting a breast from the preprocessed mammography image; and

a-3) a breast cancer risk assessment method comprising the step of calculating a breast density using the pixel density of the breast region.
According to claim 4,

In step a-2),

dividing the mammography image into a first density area, a second density area, and a third density area in descending order of brightness values according to a preset standard of brightness values for each pixel; and

Comprising the step of calculating the size of the first to third density areas of the breast region, breast cancer risk assessment method.
According to claim 5,

The size of the breast area for each of the first to third density areas,

Breast size corresponding to the absolute value of the breast size corresponding to the area value of the first to third density regions of the breast region and the value obtained by dividing the area value of the first to third density regions of the breast region by the total area value of the breast A method for assessing breast cancer risk, including relative size.
According to claim 5,

In step a-3),

Calculating breast density for each region using pixel densities in the first to third density regions.
According to claim 1,

Step a) includes a step of measuring breast density using artificial intelligence technology,

The breast density measurement step using the artificial intelligence technology,

Based on a plurality of mammography images to be learned, vector transformation to which weights are applied is performed to calculate the breast density of the breast area in the image, and the calculated breast density and the plurality of mammography images to be learned are calculated in the image derived by the expert. The step of measuring the density of the breast to be evaluated using an artificial intelligence model learned to minimize the difference in breast density in the breast area, the breast cancer risk assessment method.
According to claim 1,

Step a) includes a pattern extraction step using artificial intelligence technology,

The feature pattern includes an abnormal breast feature pattern of the breast to be evaluated,

The pattern extraction step using the artificial intelligence technology,

By using an artificial intelligence model for extracting abnormal breast patterns configured to extract characteristic patterns of abnormal breasts from the input images by performing machine learning based on mammogram images of normal people and breast cancer patients, the abnormality of the breast to be evaluated from the mammogram images. A breast cancer risk assessment method comprising the step of extracting a breast feature pattern.
In the breast cancer risk assessment system using mammography images,

a communication module receiving the mammography image;

a memory for storing a breast cancer risk assessment program; and

A processor for executing a breast cancer risk assessment program stored in the memory;

The processor executes the breast cancer risk assessment program,

Evaluation data including breast density information generated by measuring the density of the breast to be evaluated from a mammogram image of the breast to be evaluated and breast pattern information generated by extracting a characteristic pattern of the breast to be evaluated from the mammogram image Generating, and configured to calculate the risk of breast cancer occurrence for the breast to be evaluated by applying a predetermined weight to each of the information included in the evaluation data, breast cancer risk evaluation system.
According to claim 10,

The processor executes the breast cancer risk assessment program,

It is configured to further generate clinical information generated based on at least one or more of age information, genetic information, and breast cancer family history information of a specific person corresponding to the breast to be evaluated,

The evaluation data includes the clinical information, breast cancer risk assessment system.
According to claim 10,

The mammography image is a DICOM (Digital Imaging and Communications in Medicine) type image, breast cancer risk assessment system.
According to claim 10,

The processor executes the breast cancer risk assessment program,

Preprocessing is performed to adjust the mammography image to have a preset contrast and size structure, a breast portion is divided from the preprocessed mammography image to calculate the size of the breast to be evaluated, and pixel density of the breast portion is calculated. , Breast cancer risk assessment system configured to further perform calculation of breast density using.
According to claim 13,

The processor executes the breast cancer risk assessment program,

The mammography image is divided into a first density area, a second density area, and a third density area in descending order of brightness value according to a preset criterion of brightness value for each pixel, and then, first to second density areas of the breast region The breast cancer risk assessment system, configured to further perform calculating the size of the third density region.
According to claim 14,

The size of the breast area for each of the first to third density areas,

Breast size corresponding to the absolute value of the breast size corresponding to the area value of the first to third density regions of the breast region and the value obtained by dividing the area value of the first to third density regions of the breast region by the total area value of the breast Breast cancer risk rating system, including size relative.
According to claim 14,

The processor executes the breast cancer risk assessment program,

The breast cancer risk assessment system configured to further perform calculation of breast density by region using pixel densities in the first to third density regions.
According to claim 10,

The processor executes the breast cancer risk assessment program,

Based on a plurality of mammography images to be learned, a weighted vector transformation is performed to calculate the breast density of the breast part of the image, and the difference between the calculated breast density and the breast density derived by an expert for the plurality of mammography images to be learned is calculated. , Breast cancer risk assessment system configured to further perform measuring the density of the breast to be evaluated using an artificial intelligence model learned to minimize.
According to claim 10,

The feature pattern includes an abnormal breast feature pattern of the breast to be evaluated,

The processor executes the breast cancer risk assessment program,

By using an artificial intelligence model for extracting abnormal breast patterns configured to extract characteristic patterns of abnormal breasts from the input images by performing machine learning based on mammogram images of normal people and breast cancer patients, the abnormality of the breast to be evaluated from the mammogram images. A breast cancer risk assessment system, configured to further perform extracting a breast feature pattern.
A non-transitory computer-readable recording medium on which a computer program for implementing the breast cancer risk assessment method according to claim 1 is recorded.